Model benchmarks
Qwen3.6-27B-MTPLX-Optimized-Speed local LLM performance
As of July 2026, Qwen3.6-27B-MTPLX-Optimized-Speed runs at up to 31.2 tok/s for local inference (best of 5 community benchmark runs across 1 GPU).
Model size
27B
Peak speed
31.2 tok/s
Average speed
19.5 tok/s
Min memory
18.0 GB
Max context
65,536 tokens
Avg quality
72.5
Benchmark runs
5
GPUs tested
1
Quality by task
Average LLM-judged quality (0–100) with the run-to-run spread shown as a P5–P95 band, overall and for each benchmark task, across all 5 runs. The low and high columns show how much the judge’s score varies between runs, and need at least two runs to display.
| Task | P5 (low) | Avg | P95 (high) |
|---|---|---|---|
| Overall | 57.4 | 72.5 | 85.3 |
| Agent Workflow | 75.3 | 80.7 | 88.6 |
| Code Generation | 74.3 | 79.6 | 83.7 |
| Role Play & Narrative | 0.0 | 52.4 | 91.4 |
| Research & Analysis | 68.9 | 77.3 | 82.8 |
Performance by hardware and tool
Every hardware/tool/quantization combination Qwen3.6-27B-MTPLX-Optimized-Speed has been benchmarked on, ranked by peak token generation speed. Last updated July 2026.
| Hardware | Tool | Quant | Best tok/s | Avg tok/s | Memory | Context | Quality | Runs |
|---|---|---|---|---|---|---|---|---|
| Apple M5 Max | oMLX | — | 31.2 tok/s | 19.5 tok/s | 18.0 GB | 65,536 tokens | 72.5 | 5 |
Benchmark runs
All 5 Qwen3.6-27B-MTPLX-Optimized-Speed runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is Qwen3.6-27B-MTPLX-Optimized-Speed good for coding?
- In our benchmarks, Qwen3.6-27B-MTPLX-Optimized-Speed scores 79.6/100 for coding. It runs at about 19.5 tok/s, so if you want more speed, Qwen3.8-27B-Turbo-Brilliance-50X-RI-GSQ-RCO-IQ3_XXS-mtp is faster (~37.5 tok/s) and still scores well for coding (83.8/100).
- Is Qwen3.6-27B-MTPLX-Optimized-Speed good for agentic (tool-using) tasks?
- In our benchmarks, Qwen3.6-27B-MTPLX-Optimized-Speed scores 80.7/100 for agentic workflows. It runs at about 19.5 tok/s, so if you want more speed, muse-glimmer:latest is faster (~33.3 tok/s) and still scores well for agentic workflows (91.6/100).
- How fast is Qwen3.6-27B-MTPLX-Optimized-Speed for local inference?
- Across 5 community benchmark runs, Qwen3.6-27B-MTPLX-Optimized-Speed reaches up to 31.2 tok/s and averages 19.5 tok/s, with the fastest results on Apple M5 Max.
- How much memory does Qwen3.6-27B-MTPLX-Optimized-Speed need?
- The leanest observed configuration used about 18.0 GB of memory.
- Which tools have been used to run Qwen3.6-27B-MTPLX-Optimized-Speed?
- Benchmarks were submitted using oMLX. Results are community-contributed and updated as new runs arrive.